Problems subloading GL_LUMINANCE_FLOAT32_ATI textures

I’ve been working on using texture lookups in the vertex shader. I’m having trouble with subloading GL_LUMINANCE_FLOAT32_ATI textures, and was wondering if anyone else has had similar issues. I am running with a 6800 GT and have seen the results described below using both 66.93 drivers and 66.77 beta drivers.

What I am trying to do is amortize the cost of replacing the contents of a 1024x1024 displacement map. So each frame, I want the application to load a 1024x32 portion of the image.

Essentially, if I use any other internal format besides GL_LUMINANCE_FLOAT32_ATI and GL_RGBA_FLOAT32_ATI, glTexSubImage2D works fine. glTexSubImage2D works ok with the ATI formats only if I subload the entire extent of the image. If I attempt to subload a portion of the image, the application may crash, depending on the portion I attempt to subload. If it doesn’t crash, then I get a bizaar offset pattern in the resulting texture. If I switch to GL_LUMINANCE for the internal format, everything works.

FYI, filtering is set to nearest and mip-mapping is disabled. The source data is sent as GL_LUMINANCE and GL_UNSIGNED_BYTE. Any ideas?

Sounds like a driver bug to me.
There appeared a new beta driver here:

The 71.84 drivers seem to have corrected the problem. Thanks!